Spring Framework: Deep Dive

This page covers advanced Spring Framework concepts including the bean lifecycle, AOP, data access patterns, reactive programming, and batch processing.

Spring Bean Lifecycle

Understanding the bean lifecycle is crucial for optimizing resource management in large-scale applications.

Lifecycle Phases

Spring Bean Lifecycle Phase Explorer

6. Initialization Callbacks

What happens

Execute @PostConstruct annotations
Invoke InitializingBean.afterPropertiesSet()
Invoke custom init-method defined in @Bean(initMethod = "...")

Representative Method/Callback

@PostConstruct public void init() { ... }

💡 Senior Gotcha / Caution

The bean is initialized but not yet proxied for AOP annotations (like @Transactional or @Async) if they are handled in post-processors.

Lifecycle Callbacks

Callback	Mechanism	When It Runs
`@PostConstruct`	Annotation	After dependency injection is complete
`InitializingBean.afterPropertiesSet()`	Interface	After all properties are set
Custom `init-method`	XML/annotation config	After `afterPropertiesSet()`
`@PreDestroy`	Annotation	Before bean is removed from container
`DisposableBean.destroy()`	Interface	During container shutdown
Custom `destroy-method`	XML/annotation config	After `destroy()`

@Component
public class DataSourceManager {

    @PostConstruct
    public void init() {
        // Initialize connection pool
    }

    @PreDestroy
    public void cleanup() {
        // Close connections gracefully
    }
}

ApplicationContext vs BeanFactory

Feature	BeanFactory	ApplicationContext
Bean Instantiation	Lazy (on demand)	Eager (at startup)
Event Propagation	No	Yes (`ApplicationEvent`)
AOP Integration	Manual	Built-in
Internationalization (i18n)	No	Yes (`MessageSource`)
Web Context Support	No	Yes (`WebApplicationContext`)
Resource Loading	Basic	Advanced (`ResourceLoader`)
Recommended For	Low-memory / embedded systems	Enterprise applications

// BeanFactory (basic)
BeanFactory factory = new XmlBeanFactory(new ClassPathResource("beans.xml"));

// ApplicationContext (preferred)
ApplicationContext context = new AnnotationConfigApplicationContext(AppConfig.class);

BeanPostProcessor vs BeanFactoryPostProcessor

Two critical extension points that senior engineers must distinguish:

	`BeanFactoryPostProcessor`	`BeanPostProcessor`
When it runs	Before any beans are instantiated	After each bean is instantiated
What it modifies	Bean definitions (metadata)	Bean instances
`ApplicationContext.getBean()` safe?	No — triggers premature instantiation	Yes
Common use	`PropertySourcesPlaceholderConfigurer` resolving `${...}`	AOP proxying, `@Autowired` injection

// BeanFactoryPostProcessor — modifies bean definitions before instantiation
@Component
public class CustomBeanFactoryPostProcessor implements BeanFactoryPostProcessor {
    @Override
    public void postProcessBeanFactory(ConfigurableListableBeanFactory factory) {
        BeanDefinition def = factory.getBeanDefinition("myService");
        def.setScope(BeanDefinition.SCOPE_PROTOTYPE);  // Change scope at definition level
    }
}

// BeanPostProcessor — wraps/modifies bean instances after creation
@Component
public class AuditBeanPostProcessor implements BeanPostProcessor {
    @Override
    public Object postProcessAfterInitialization(Object bean, String beanName) {
        if (bean instanceof AuditableService) {
            return Proxy.newProxyInstance(...)  // Wrap in dynamic proxy
        }
        return bean;
    }
}

Interview trap: BeanPostProcessor beans are instantiated very early in the context lifecycle, before other beans. Never @Autowire a late-initializing bean (e.g., JPA repositories) into a BeanPostProcessor — it causes premature initialization and bypasses auto-configuration.

Circular Dependencies

A circular dependency occurs when two or more beans depend on each other:

Bean A → requires Bean B → requires Bean A → deadlock!

Resolution Strategies

Strategy	How It Works
Setter Injection	Allows beans to be instantiated before dependencies are set
`@Lazy` Annotation	Defers bean initialization until actually needed, breaking the cycle
Redesign Architecture	Introduce an interface or third bean to decouple

@Component
public class ServiceA {
    private ServiceB serviceB;

    @Autowired
    @Lazy
    public void setServiceB(ServiceB serviceB) {
        this.serviceB = serviceB;
    }
}

Stereotype Annotations

Annotation	Layer	Purpose
`@Component`	Generic	Any Spring-managed component
`@Service`	Service	Business logic and service tasks
`@Repository`	Data Access	Database interaction, exception translation
`@Controller`	Presentation	Web request handling (MVC)
`@RestController`	Presentation	RESTful web services (`@Controller` + `@ResponseBody`)

@Component, @Service, @Repository, and @Controller are technically interchangeable — they all register beans. However, using the correct stereotype improves code clarity and enables layer-specific features (e.g., @Repository adds persistence exception translation).

Data Access: JpaRepository vs CrudRepository

Feature	CrudRepository	JpaRepository
CRUD Operations	Yes	Yes (inherited)
Pagination & Sorting	No	Yes
Batch Operations	No	Yes (`saveAll`, `deleteInBatch`)
Flush Persistence Context	No	Yes (`flush()`, `saveAndFlush()`)
Best For	Simple data access	Full JPA capabilities

// CrudRepository — basic CRUD
public interface UserRepository extends CrudRepository<User, Long> {
}

// JpaRepository — full JPA features
public interface OrderRepository extends JpaRepository<Order, Long> {
    List<Order> findByStatus(OrderStatus status);
}

@Qualifier vs @Primary

When multiple beans of the same type exist, Spring needs to know which one to inject.

@Configuration
public class DataSourceConfig {

    @Bean
    @Primary  // Default choice when no qualifier specified
    public DataSource primaryDataSource() {
        return new HikariDataSource(primaryConfig());
    }

    @Bean
    @Qualifier("reporting")
    public DataSource reportingDataSource() {
        return new HikariDataSource(reportingConfig());
    }
}

@Service
public class ReportService {
    // Uses the @Qualifier to pick a specific bean
    public ReportService(@Qualifier("reporting") DataSource dataSource) {
        // ...
    }
}

Annotation	Behavior
`@Primary`	Marks a bean as the default when multiple candidates exist
`@Qualifier`	Explicitly selects a specific bean by name

@Transactional

The @Transactional annotation defines the scope of a database transaction. All operations within the annotated method either succeed or fail together.

@Service
public class TransferService {

    @Transactional
    public void transfer(Long fromId, Long toId, BigDecimal amount) {
        accountRepository.debit(fromId, amount);
        accountRepository.credit(toId, amount);
        // If credit fails, debit is rolled back
    }
}

Key Attributes

Attribute	Purpose	Default
`propagation`	How transactions relate to each other	`REQUIRED`
`isolation`	Transaction isolation level	Database default
`readOnly`	Hint for optimization on read-only operations	`false`
`rollbackFor`	Exceptions that trigger rollback	Unchecked exceptions
`timeout`	Maximum time for the transaction	No timeout

@Transactional Pitfall Patterns

These are the most common senior interview and production bug topics:

1. Checked exceptions don't roll back by default

@Transactional
public void processPayment(Order order) throws PaymentException {
    paymentRepository.save(order);
    paymentGateway.charge(order);  // Throws PaymentException (checked)
    // ❌ Transaction COMMITS even though exception was thrown!
    // PaymentException is a checked exception — not rolled back by default
}

// Fix:
@Transactional(rollbackFor = PaymentException.class)
public void processPayment(Order order) throws PaymentException { ... }

2. @Transactional on private or final methods is silently ignored

@Service
public class OrderService {
    @Transactional  // ❌ IGNORED — CGLIB cannot override private/final methods
    private void saveOrder(Order order) { ... }
}
// Solution: always put @Transactional on public methods

3. Transaction not applied to new threads

@Transactional
public void processOrders(List<Order> orders) {
    orders.parallelStream().forEach(order -> {
        orderRepository.save(order);  // ❌ No transaction! Runs in a new thread
    });
}
// The transaction context is bound to the current thread via ThreadLocal.
// New threads spawned inside a @Transactional have NO transaction.

4. REQUIRES_NEW doesn't work when called internally (self-invocation)

@Transactional
public void outer() {
    inner();  // ❌ @Transactional(REQUIRES_NEW) on inner() is IGNORED via self-call
}

@Transactional(propagation = Propagation.REQUIRES_NEW)
public void inner() { ... }  // Only works if called from ANOTHER Spring bean

5. readOnly = true is not enforced — it's a hint

@Transactional(readOnly = true)
public User getUser(Long id) {
    User user = userRepository.findById(id).orElseThrow();
    user.setName("Modified");  // Hibernate may or may not flush this — behavior is DB/driver dependent
    return user;
    // readOnly = true skips the "dirty check" — Hibernate won't check for changes
    // This is a performance optimization, NOT a write-guard
}

Aspect-Oriented Programming (AOP) — Deep Dive

AOP modularizes cross-cutting concerns — functionality that spans multiple classes like logging, security, and transaction management — keeping business logic clean.

Core Concepts

Spring AOP Proxy Interception Architecture

Caller Client (User Thread)

▼ invokes method

Spring AOP Proxy Wrapper

CGLIB Subclass Proxy / JDK Proxy

Advice Interceptor (@Before, @Around)

▼ delegates on proceed()

Target Object (Raw Bean Entity)

2. Spring AOP Proxy

Architecture Role

Generated wrapper (CGLIB subclass or JDK Dynamic Proxy). Intercepts all incoming invocations on public methods.

⚠️ Critical Gotcha / Trap

Cannot proxy final classes, final methods, or private methods. Forces CGLIB by default in Spring Boot 2.x/3.x.

Spring AOP vs AspectJ

	Spring AOP	AspectJ
Weaving	Proxy-based (runtime)	Bytecode weaving (compile/load time)
Join points supported	Method execution only	Fields, constructors, static methods, etc.
Self-invocation	❌ (bypasses proxy)	✅ (weaved into bytecode)
Setup	Zero config (built into Spring)	Needs AspectJ compiler / agent
Best for	Most enterprise use cases	Performance-critical or non-Spring code

Key insight: Spring AOP only intercepts Spring-managed bean method calls made through the proxy. Field access, constructors, and this.method() calls are invisible to it.

Pointcut Expression DSL

Mastery of pointcut expressions is essential for writing precise, production-safe aspects:

// execution — most common: matches method execution
@Pointcut("execution(* com.example.service.*.*(..))")
// execution( [modifier] [return-type] [declaring-type.][method]([params]) )
//   *            = any return type
//   com.example.service.*.*(..)  = any method in any class in this package
//   (..) = any number of params

// within — matches all methods within a type/package
@Pointcut("within(com.example.controller..*)")  // all classes in controller and subpackages

// @annotation — matches methods annotated with a specific annotation
@Pointcut("@annotation(com.example.annotation.Auditable)")

// @within — matches all methods within a class annotated with a given annotation
@Pointcut("@within(org.springframework.stereotype.Service)")

// bean — matches methods on specific Spring beans (Spring AOP only)
@Pointcut("bean(orderService)")
@Pointcut("bean(*Service)")  // all beans ending in "Service"

// args — matches based on method argument types at runtime
@Pointcut("args(com.example.dto.OrderRequest, ..)")

// Combining expressions
@Pointcut("within(com.example.service..*) && !execution(* *.get*(..)))")
// all service methods except getters

Advice Types

@Aspect
@Component
public class ExampleAspect {

    // Runs before the method — cannot prevent execution (use @Around for that)
    @Before("execution(* com.example.service.*.*(..))")
    public void beforeAdvice(JoinPoint jp) {
        log.info("Calling: {}", jp.getSignature().toShortString());
    }

    // Runs after normal return — receives the return value
    @AfterReturning(pointcut = "execution(* com.example.service.*.*(..))", returning = "result")
    public void afterReturning(JoinPoint jp, Object result) {
        log.info("Returned: {}", result);
    }

    // Runs after an exception is thrown
    @AfterThrowing(pointcut = "execution(* com.example.service.*.*(..))", throwing = "ex")
    public void afterThrowing(JoinPoint jp, Exception ex) {
        log.error("Exception in {}: {}", jp.getSignature(), ex.getMessage());
    }

    // Runs after method regardless of outcome (like finally)
    @After("execution(* com.example.service.*.*(..))")
    public void afterAdvice(JoinPoint jp) { }

    // Wraps the method — full control, must call proceed()
    @Around("execution(* com.example.service.*.*(..))")
    public Object aroundAdvice(ProceedingJoinPoint pjp) throws Throwable {
        long start = System.currentTimeMillis();
        try {
            Object result = pjp.proceed();  // MUST call proceed() to continue
            log.info("Duration: {}ms", System.currentTimeMillis() - start);
            return result;
        } catch (Exception e) {
            log.error("Failed after {}ms", System.currentTimeMillis() - start);
            throw e;  // Always rethrow unless intentionally swallowing
        }
    }
}

Aspect Ordering (`@Order`)

When multiple aspects apply to the same join point, @Order controls execution order. Lower number = outer (runs first entering, last exiting):

@Order(1) Security Aspect → @Order(2) Logging Aspect → @Order(3) Transaction Aspect
         ↓                           ↓                           ↓
                              Target Method
         ↑                           ↑                           ↑
    (last out)                 (second out)                (first out)

@Aspect @Component @Order(1)
public class SecurityAspect { ... }  // Outermost — checks auth first

@Aspect @Component @Order(2)
public class LoggingAspect { ... }   // Middle — logs method entry/exit

@Aspect @Component @Order(3)
public class TransactionAspect { ... } // Innermost — closest to target

In Spring's built-in aspects: @Transactional has Integer.MAX_VALUE order (innermost). @Validated has Integer.MAX_VALUE - 1. Place your custom aspects with explicit lower numbers to run before them.

AOP Integration 1: Logging and MDC Tracing

The most valuable production AOP use case — attaching trace IDs, user context, and timing to every log line without cluttering service code:

@Aspect
@Component
@Order(2)
public class LoggingAspect {

    @Around("@within(org.springframework.stereotype.Service)")
    public Object logAndTrace(ProceedingJoinPoint pjp) throws Throwable {
        String method = pjp.getSignature().toShortString();
        String traceId = UUID.randomUUID().toString().substring(0, 8);

        // MDC: attach context to every log line in this thread
        MDC.put("traceId", traceId);
        MDC.put("method", method);
        long start = System.currentTimeMillis();

        try {
            log.debug("→ {}", method);
            Object result = pjp.proceed();
            log.debug("← {} [{}ms]", method, System.currentTimeMillis() - start);
            return result;
        } catch (Exception e) {
            log.error("✗ {} failed [{}ms]: {}", method, System.currentTimeMillis() - start, e.getMessage());
            throw e;
        } finally {
            MDC.clear();  // CRITICAL: always clear MDC in finally to avoid leaks in thread pools
        }
    }
}

Logback pattern to use the MDC context:

<pattern>%d{HH:mm:ss} [%thread] %X{traceId} %-5level %logger - %msg%n</pattern>

Production refinement — propagate MDC to async threads:

// MDC is ThreadLocal — it's lost when @Async spawns a new thread
// Fix: use MDC.getCopyOfContextMap() and restore in the async thread
@Async
public void asyncTask() {
    Map<String, String> context = MDC.getCopyOfContextMap();
    CompletableFuture.runAsync(() -> {
        if (context != null) MDC.setContextMap(context);
        try { doWork(); } finally { MDC.clear(); }
    });
}

AOP Integration 2: Security with `@PreAuthorize` and Method Security

Spring Security's method-level security (@PreAuthorize, @PostAuthorize, @Secured) is implemented as an AOP aspect:

@Service
public class OrderService {

    // SpEL expression evaluated against SecurityContext
    @PreAuthorize("hasRole('ORDER_MANAGER') or #order.ownerId == authentication.principal.id")
    public void cancelOrder(Order order) { ... }

    // @PostAuthorize: runs AFTER method, can inspect return value
    @PostAuthorize("returnObject.ownerId == authentication.principal.id")
    public Order getOrder(Long orderId) { ... }

    // @PostFilter: filters collection return values
    @PostFilter("filterObject.status != 'CONFIDENTIAL' or hasRole('ADMIN')")
    public List<Order> getAllOrders() { ... }
}

How it works internally:

External call → MethodSecurityInterceptor (AOP @Around advice)
                    ↓
            Evaluates SpEL against SecurityContextHolder.getContext()
                    ↓
            AccessDecisionManager → grants or throws AccessDeniedException
                    ↓
            pjp.proceed() → target method executes

Custom security aspect — audit sensitive data access:

@Aspect
@Component
@Order(1)  // Before transaction opens (so we can audit before any DB write)
public class AuditSecurityAspect {

    private final AuditRepository auditRepo;

    @Around("@annotation(com.example.annotation.SensitiveOperation)")
    public Object auditSensitiveOperation(ProceedingJoinPoint pjp) throws Throwable {
        Authentication auth = SecurityContextHolder.getContext().getAuthentication();
        String username = auth != null ? auth.getName() : "anonymous";
        String operation = pjp.getSignature().toShortString();

        AuditEntry entry = AuditEntry.builder()
            .username(username)
            .operation(operation)
            .args(Arrays.toString(pjp.getArgs()))
            .timestamp(Instant.now())
            .build();

        try {
            Object result = pjp.proceed();
            entry.setOutcome("SUCCESS");
            auditRepo.save(entry);
            return result;
        } catch (AccessDeniedException e) {
            entry.setOutcome("ACCESS_DENIED");
            auditRepo.save(entry);
            throw e;
        } catch (Exception e) {
            entry.setOutcome("ERROR: " + e.getMessage());
            auditRepo.save(entry);
            throw e;
        }
    }
}

Common pitfall: @PreAuthorize on a Spring @Component method works. @PreAuthorize on a controller method called internally (this.method()) does NOT — same self-invocation problem as @Transactional.

AOP Integration 3: Caching with `@Cacheable`

Spring's @Cacheable is an AOP around-advice that:

Computes the cache key from SpEL
Checks the cache — returns cached value if hit
Calls the target method on cache miss
Stores the result in the cache

@Service
public class ProductService {

    // key is evaluated from method args using SpEL
    @Cacheable(value = "products", key = "#category + '_' + #page", unless = "#result.isEmpty()")
    public List<Product> getProductsByCategory(String category, int page) {
        return productRepository.findByCategory(category, PageRequest.of(page, 20));
    }

    // @CacheEvict: removes entries on write operations
    @CacheEvict(value = "products", allEntries = true)
    @Transactional
    public Product updateProduct(Product product) {
        return productRepository.save(product);
    }

    // @CachePut: always updates cache regardless of hit/miss
    @CachePut(value = "products", key = "#result.id")
    public Product createProduct(Product product) {
        return productRepository.save(product);
    }
}

Production gotchas:

// ❌ Self-invocation — @Cacheable is ignored (same proxy problem)
public List<Product> getAll() {
    return getProductsByCategory("electronics", 0);  // NOT cached!
}

// ❌ @Cacheable on private/final method — silently ignored
@Cacheable("products")
private List<Product> findFromDb() { ... }  // CGLIB can't intercept

// ❌ Cache key collision — different methods, same key expression
@Cacheable(value = "users", key = "#id")    // User cache
@Cacheable(value = "users", key = "#id")    // Order cache (different bean, same value!)
// Fix: always use distinct cache value names per entity type

Custom cache aspect for metrics:

@Aspect
@Component
@Order(10)  // After Spring's @Cacheable aspect (which has lower order)
public class CacheMetricsAspect {

    private final MeterRegistry meterRegistry;

    @Around("@annotation(cacheable)")
    public Object trackCacheUsage(ProceedingJoinPoint pjp, Cacheable cacheable) throws Throwable {
        Timer.Sample sample = Timer.start(meterRegistry);
        String cacheName = cacheable.value()[0];

        Object result = null;
        boolean cacheHit = false;

        result = pjp.proceed();
        // Spring's cacheable aspect has already run — if result was cached, method wasn't called
        // We'd need to hook CacheManager directly to know hit/miss precisely

        sample.stop(meterRegistry.timer("cache.operation", "cache", cacheName));
        return result;
    }
}

AOP Integration 4: Transactions — How `@Transactional` Uses AOP

@Transactional is an AOP around-advice implemented by TransactionInterceptor:

External call
    → CGLIB/JDK Proxy (TransactionInterceptor's @Around)
        → TransactionManager.getTransaction() (opens DB connection + begins TX)
            → target method executes (JPA calls, etc.)
        → on success: TransactionManager.commit()
        → on exception: TransactionManager.rollback() (for unchecked + configured exceptions)

Transaction propagation modeled as AOP advice chain:

// Service A @Transactional (REQUIRED) calls Service B @Transactional (REQUIRES_NEW)
// These are separate Spring beans — TWO separate proxy invocations:

class ServiceA {                          class ServiceB {
  TX-Proxy-A                              TX-Proxy-B
    @Transactional REQUIRED                 @Transactional REQUIRES_NEW
    ↓                                       ↓
  Begin TX-1                              Suspend TX-1
    → calls ServiceB.save()               Begin TX-2
       → ServiceB proxy intercepts           → execute
          → REQUIRES_NEW → suspend TX-1   Commit TX-2
                                          Resume TX-1
}                                       → Return to ServiceA

Thread boundary: transactions don't cross @Async barriers

@Service
public class ReportService {

    @Transactional
    public void generateReport() {
        List<Order> orders = orderRepo.findAll();

        // ❌ @Async spawns new thread — transaction context is ThreadLocal, NOT transferred
        notificationService.sendReport(orders);
        // notificationService.sendReport() runs in a thread with NO transaction
    }
}

AOP Integration 5: Metrics and Monitoring with Micrometer

AOP makes it trivial to add performance metrics without modifying business code:

@Target(ElementType.METHOD)
@Retention(RetentionPolicy.RUNTIME)
public @interface Timed {
    String value() default "";
    String[] tags() default {};
}

@Aspect
@Component
public class MetricsAspect {

    private final MeterRegistry meterRegistry;

    @Around("@annotation(timed)")
    public Object recordMetrics(ProceedingJoinPoint pjp, Timed timed) throws Throwable {
        String metricName = timed.value().isEmpty()
            ? pjp.getSignature().getDeclaringTypeName() + "." + pjp.getSignature().getName()
            : timed.value();

        Timer.Sample sample = Timer.start(meterRegistry);

        try {
            Object result = pjp.proceed();
            sample.stop(meterRegistry.timer(metricName,
                "outcome", "success",
                "class", pjp.getSignature().getDeclaringType().getSimpleName()));
            return result;
        } catch (Exception e) {
            sample.stop(meterRegistry.timer(metricName,
                "outcome", "error",
                "exception", e.getClass().getSimpleName()));
            meterRegistry.counter(metricName + ".errors",
                "exception", e.getClass().getSimpleName()).increment();
            throw e;
        }
    }
}

// Usage on service methods
@Service
public class OrderService {
    @Timed("order.placement")
    public Order placeOrder(OrderRequest req) { ... }
}

Integrating with Micrometer's @Timed directly:

// Spring Boot Actuator + Micrometer already provides @io.micrometer.core.annotation.Timed
// Enable via TimedAspect bean:
@Bean
public TimedAspect timedAspect(MeterRegistry registry) {
    return new TimedAspect(registry);
}

// Then on any Spring bean method:
@Timed(value = "http.requests", extraTags = {"region", "us-east-1"}, percentiles = {0.5, 0.95, 0.99})
public Order placeOrder(OrderRequest req) { ... }
// Metrics available at /actuator/metrics/http.requests

AOP Integration 6: Retry and Circuit Breaker

Both Spring Retry and Resilience4j use AOP aspects under the hood:

Spring Retry (@Retryable):

@Service
public class PaymentService {

    // AOP intercepts this → retries on failure with backoff
    @Retryable(
        retryFor = {TransientDataAccessException.class, RemoteServiceException.class},
        maxAttempts = 3,
        backoff = @Backoff(delay = 1000, multiplier = 2.0, maxDelay = 10000)
    )
    public PaymentResult charge(PaymentRequest request) {
        return paymentGateway.process(request);
    }

    @Recover  // Called when all retries exhausted
    public PaymentResult chargeRecover(RemoteServiceException ex, PaymentRequest request) {
        log.error("Payment failed after retries: {}", request.getId());
        return PaymentResult.failed("service_unavailable");
    }
}

Resilience4j with AOP:

@Service
public class InventoryService {

    // Multiple AOP aspects applied — order matters!
    // Execution order (inner to outer):
    // Bulkhead → Rate Limiter → Circuit Breaker → Retry → TimeLimiter → method
    @CircuitBreaker(name = "inventory", fallbackMethod = "fallbackInventory")
    @Retry(name = "inventory")
    @TimeLimiter(name = "inventory")
    public CompletableFuture<Integer> checkStock(Long productId) {
        return CompletableFuture.supplyAsync(() -> inventoryClient.getStock(productId));
    }

    // Fallback receives the exception as first param
    public CompletableFuture<Integer> fallbackInventory(Long productId, CallNotPermittedException ex) {
        log.warn("Circuit open for product {}: using cached stock", productId);
        return CompletableFuture.completedFuture(stockCache.getOrDefault(productId, 0));
    }
}

Resilience4j is NOT a Spring AOP aspect by default — it has its own annotation-based AOP that requires resilience4j-spring-boot3 starter. The proxy chain stacks multiple Resilience4j interceptors.

Custom Annotation-Based Aspect Pattern

The cleanest pattern: define your own annotation, then intercept it with an aspect — no pointcut expressions in business code:

// 1. Define the annotation
@Target(ElementType.METHOD)
@Retention(RetentionPolicy.RUNTIME)
@Documented
public @interface RateLimit {
    int requestsPerSecond() default 100;
    String key() default "";  // SpEL for dynamic key (e.g., "#userId")
}

// 2. Binding annotation to advice via parameter binding
@Aspect
@Component
public class RateLimitAspect {

    private final RateLimiterRegistry registry = RateLimiterRegistry.ofDefaults();

    @Around("@annotation(rateLimit)")  // Binds annotation instance to param
    public Object enforceRateLimit(ProceedingJoinPoint pjp, RateLimit rateLimit) throws Throwable {
        // Parse SpEL key expression
        String key = resolveKey(rateLimit.key(), pjp);

        RateLimiter limiter = registry.rateLimiter(key,
            RateLimiterConfig.custom()
                .limitForPeriod(rateLimit.requestsPerSecond())
                .limitRefreshPeriod(Duration.ofSeconds(1))
                .build());

        return limiter.executeCallable(() -> {
            try { return pjp.proceed(); }
            catch (Throwable t) { throw new RuntimeException(t); }
        });
    }
}

// 3. Apply cleanly in service code
@Service
public class ApiService {
    @RateLimit(requestsPerSecond = 50, key = "#userId")
    public Response handleRequest(Long userId, Request req) { ... }
}

AOP Aspect Ordering — Spring's Internal Stack

When multiple framework annotations combine, the proxy chain from outermost to innermost looks like this in a typical Spring Boot service call:

Spring AOP Aspect Ordering Proxy Chain

REQUEST FLOW (Top Down) ──► RESPONSE FLOW (Bottom Up)

Spring Security (@PreAuthorize)Order 1

Logging / Tracing (MDC setup)Order 2

Metrics Aspect (@Timed)Order 3

Retry Aspect (@Retryable)Order 4

Cache Aspect (@Cacheable)Order 5

Transaction Aspect (@Transactional)Order MAX-1

Validation Aspect (@Validated)Order MAX

🎯 Target Method Execution (Inner Business Logic)

Cache Aspect (@Cacheable)Precedence Chain: Order 5

Execution Role

Intercepts call, checks cache manager, returns cached item immediately if present.

⚠️ Critical Order Gotcha

Must sit OUTSIDE @Transactional. Caching reads should avoid opening transaction connections to free up database resources.

Understanding this chain is critical for debugging puzzling behavior like:

Why did my audit log save even though the transaction rolled back? (Audit aspect is outside transaction)
Why is my cache returning stale data after a failed update? (CacheEvict runs even on exception by default — use beforeInvocation = false)
Why isn't my @Retryable working? (It's on a private method or called internally)

Same issue applies to: @Async, @Cacheable, @Scheduled, and any other Spring AOP annotation.

Spring WebFlux vs Spring MVC

Aspect	Spring MVC	Spring WebFlux
Programming Model	Synchronous, blocking	Asynchronous, non-blocking
Concurrency	Thread-per-request	Event-loop (fewer threads)
Built On	Servlet API	Project Reactor
Best For	Traditional web apps	High-concurrency, streaming
Server	Tomcat, Jetty	Netty, Undertow

// Spring MVC (blocking)
@GetMapping("/users/{id}")
public User getUser(@PathVariable Long id) {
    return userService.findById(id);
}

// Spring WebFlux (reactive)
@GetMapping("/users/{id}")
public Mono<User> getUser(@PathVariable Long id) {
    return userService.findById(id);
}

Spring Batch

Spring Batch is a framework for processing large volumes of data efficiently — ideal for data migration, report generation, and scheduled jobs.

Architecture

Spring Batch Architecture & Chunk Processing Engine

Spring Batch Core Architecture & JobRepository State Machine

A JobLauncher starts a Job with unique JobParameters. A Job contains one or more Steps. All execution states (commits, skips, rollback points) are persisted atomically to the JobRepository (tables: BATCH_JOB_INSTANCE, BATCH_JOB_EXECUTION, BATCH_STEP_EXECUTION).

Key Components

Component	Role
Job	Defines the entire batch process
Step	A single phase within a job
ItemReader	Reads input data
ItemProcessor	Transforms data
ItemWriter	Writes output data
JobRepository	Stores metadata about job executions

Testing: @Mock vs @Spy

Annotation	Behavior	Use Case
`@Mock`	Fully mocked instance; no real code executes	Isolating dependencies in unit tests
`@Spy`	Partial mock wrapping a real instance; real methods execute unless overridden	Testing with some real behavior

@ExtendWith(MockitoExtension.class)
class OrderServiceTest {

    @Mock
    private PaymentGateway paymentGateway; // Fully mocked

    @Spy
    private OrderValidator orderValidator; // Real logic, selectively stubbed

    @InjectMocks
    private OrderService orderService;

    @Test
    void shouldProcessOrder() {
        when(paymentGateway.charge(any())).thenReturn(true);
        doReturn(true).when(orderValidator).validate(any()); // Override one method

        orderService.process(new Order());
        verify(paymentGateway).charge(any());
    }
}

Configuration: Annotations vs XML

Aspect	Annotations	XML
Readability	Concise, inline with code	Verbose, separate files
Maintenance	Easier — part of the codebase	Harder — separate from code
Flexibility	Requires recompilation for changes	Can be modified without recompilation
Complex Config	Can get cluttered	Better for complex wiring
Best For	Most modern projects	Legacy systems, external config needs

Best practice: Use annotations for most configurations. Reserve XML for cases where external configuration without recompilation is required.

Auto-Configuration Conflicts

When multiple @AutoConfiguration classes define the same bean, the last one loaded takes precedence. Control ordering with:

Annotation	Purpose
`@AutoConfigureOrder`	Set explicit ordering priority
`@AutoConfigureAfter`	Load after a specific auto-configuration
`@AutoConfigureBefore`	Load before a specific auto-configuration
`@ConditionalOnMissingBean`	Only create bean if it doesn't already exist

@AutoConfiguration
@AutoConfigureAfter(DataSourceAutoConfiguration.class)
public class CustomDataSourceConfig {

    @Bean
    @ConditionalOnMissingBean
    public DataSource dataSource() {
        return new CustomDataSource();
    }
}

Advanced Editorial Pass: Deep Dive with Operability Focus

Advanced Lens

Internal extension points should be used sparingly and with clear ownership.
Container behavior must remain explainable to on-call engineers under stress.
Framework customization is justified only when it reduces net complexity.

Failure Scenarios

Custom post-processors that alter bean semantics unexpectedly.
Complex proxy stacks that blur transaction and security boundaries.
Hard-to-reproduce context initialization issues across environments.

Implementation Guidance

Document every non-default extension with intent and rollback approach.
Keep AOP and proxy layering transparent in diagnostics.
Add minimal reproducible tests for every lifecycle customization.

Compare Next

Interview Questions

Q: How do you decide whether a cross-cutting concern belongs in AOP?

A: Use AOP for orthogonal policies such as logging, security, metrics, and transactions, not core business branching.

Q: What is the highest-impact proxy pitfall in Spring services?

A: Self-invocation bypasses proxies, which can silently disable @Transactional, @Async, and @Cacheable behavior.

Q: Why should bean post-processing be handled carefully?

A: Early lifecycle hooks can trigger premature bean creation and unstable startup order.

Q: How do you make transaction boundaries reliable in large codebases?

A: Keep transactional entry points explicit, public, and close to use-case orchestration boundaries.

Q: What does senior-level AOP debugging look like?

A: Trace advisor order, proxy type, and join-point matching before changing business code.

Q: When should you avoid creating another custom aspect?

A: When framework-provided mechanisms already cover the concern with lower complexity.

Q: How do method security and transaction aspects interact operationally?

A: Aspect order determines whether access checks happen before resource usage and transaction opening.